10/2/2017

1

Geography 38/42:286

GIS 1

Topic 5:

Raster and Vector Data Models

Chapters 3 & 4: Chang(Chapter 4: DeMers)

1

The Nature of Geographic Data

Most features or phenomena occur as either:

discrete entities (objects/PLPs)

continuously varying phenomena (fields/CS)

some could go either way

2

GIS Data Representation

How PLPs and CS are stored

Two primary GIS data models

Raster

Vector

3

10/2/2017

2

Data Structures and Data Models

“Raster” and “Vector” refer to a particular data model

A data structure refers to a particular implementation of

either the raster or vector model

4

Topology

The spatial relationships between things

An important distinction between data models and data structures

Can be stored or calculated “on the fly”

5

Raster – Vector Data Models

Numerous differences in terms of:accuracy

storage

functionality

efficiency

topology

representation of disrete vs continuously distributed phenomena

6

10/2/2017

3

Database Models

Also differences in how attribute data are stored

7

Raster Data Model

Study area divided up into a regular array of

discrete grid cells of uniform shape and size

8

Raster Representation of PLPs

73

Figure 5.2

Representation of point, line,

and area features: raster

format on the left and vector

format on the right.

9

10/2/2017

4

Sources of Raster Data

Common raster data sources include:

Imagery (satellite, aerial photography)

Classified imagery

DEMs, DSMs, DTMs, etc.

Scanned products (e.g. maps)

Interpolated or generated “surfaces"

10

Thematic vs. Discrete

Thematic rasters - discrete representations

Continuous rasters - continuous representations

11

Accuracy of Raster Data

Determined by grid cell size

Referred to as resolution

User defined

Sensor design

12

10/2/2017

5

Projecting Raster Data

Grid cells represent same location on the ground

Georeferencing aligns raster layer to real world

coordinate system

13

14

Advantages of Raster Model

Advantages:

Easy to understand

Mathematical operations

Representing continuous phenomena

15

10/2/2017

6

Disadvantages of Raster Model

Disadvantages:Representing discrete phenomena

Data redundancy

No topology (for discrete data)

Spatial accuracy

16

Raster Data Structures

Method by which raster model is implemented in a

particular GIS software application

Typically, user is unaware of how data structure works

17

79

Figure 5.8

The cell-by-cell data structure

records each cell value by row

and column.

Cell by Cell

18

10/2/2017

7

80

Figure 5.9

The run length encoding method

records the cell values in runs. Row

1, for example, has two adjacent

cells in columns 5 and 6 that are

gray or have the value of 1. Row 1

is therefore encoded with one run,

beginning in column 5 and ending in

column 6. The same method is used

to record other rows.

Run Length Encoding8,8,1

0,4,1,2,0,2

0,3,1,3,0,2

0,2,1,5,0,1

0,2,1,5,0,1

0,2,1,5,0,1

0,1,1,6,0,1

0,1,1,6,0,1

0,8

19

80

Figure 5.9

The run length encoding method

records the cell values in runs. Row

1, for example, has two adjacent

cells in columns 5 and 6 that are

gray or have the value of 1. Row 1

is therefore encoded with one run,

beginning in column 5 and ending in

column 6. The same method is used

to record other rows.

Block Codes

1,1,2,4,6,2,7,2

4,1,1,5

25,1,3,3

20

80

Figure 5.9

The run length encoding method

records the cell values in runs. Row

1, for example, has two adjacent

cells in columns 5 and 6 that are

gray or have the value of 1. Row 1

is therefore encoded with one run,

beginning in column 5 and ending in

column 6. The same method is used

to record other rows.

Chain Codes

0=N, 1=E, 2=S, 3=W

1,5,1

1,1,2,2,1,1,2,4,3,5,0,1,1,1,0,3

1,1,0,1,1,1,0,1

21

10/2/2017

8

81

Figure 5.10

The regional quad tree

method divides a raster into a

hierarchy of quadrants. The

division stops when a

quadrant is made of cells of

the same value (gray or

white). A quadrant that

cannot be subdivided is

called a leaf node. In the

diagram, the quadrants are

indexed spatially: 0 for NW, 1

for SW, 2 for SE, and 3 for

NE. Using the spatial

indexing method and the

hierarchical quad tree

structure, the gray cells can

be coded as 02, 032, and so

on. See text for more

explanation.

0

1 2

30

1 2

30

1 2

3

Quad Tree

22

82

Figure 5.11

The Haar wavelet and the wavelet transform. (a) Three Haar wavelets at

three scales (resolutions). (b) A simple example of the wavelet transform.

Mr SID

23

Vector Data Model

Study area is referenced to a coordinate system

that is used to identify the location of points,

lines, and polygons using x, y, (z) coordinates

24

10/2/2017

9

Vector Representation of PLPs

25

Vector Data Sources

Common vector data sources include:

NTDB data (National Topographic Database)

MLI data (1:20k roads, hydrology, etc.)

Tiger Data (U.S. Cesus)

GPS or Total Station survey data

Digitized data

26

Vector Data Model

Advantages:

Representation of discrete objects

No data redundancy

Spatial accuracy

Topology

27

10/2/2017

10

Vector Data Model

Disadvantages:

Mathematical analysis

Not suitable for continuously distributed data

28

Vector Data Models

Non-topological

Topological

29

Vector Data Models

Today there are two basic vector data models

Georelational Vector Data Model

– Arc/Info coverages & ArcView shapefiles

Object-based Vector Data Model

– ArcGIS geodatabases

30

10/2/2017

11

40

Figure 3.2

Based on the georelational data model, an ArcInfo coverage

has two components: graphic files for spatial data and INFO

files for attribute data. The label connects the two components.

Georelational

Data Model

31

45

Figure 3.7

The data structure of a point

coverage.

32

46

Figure 3.8

The data structure of a line coverage.

33

10/2/2017

12

47

Figure 3.9

The data structure of a polygon coverage.

34

48

Figure 3.10

A polygon coverage, shown in a, has topological errors. Each small square

symbol represents an error caused by lines that do not meet correctly. The

shapefile, shown in b, is converted from the polygon coverage.

35

Georelational Data Model

ArcView shapefiles

Non-topological vector data model

Topology on the fly

Advantages

36

10/2/2017

13

37

Object-based Model

Based on object oriented programming theory

Objects:

represent features

have defined properties and methods

can be organized into classes

38

Geodatabase Model

Geospatial features represented using objects

Features have properties and methods

Features and attributes stored in single file/database

39

10/2/2017

14

78

Figure 4.1

The object-based data model stores each land use polygon in a record.

The Shape field stores the spatial data of land use polygons. Other fields

store attribute data such as Landuse_ID and Category.

Binary field storing feature geometry

40

89

Figure 4.8

In a geodatabase, feature classes can be standalone

feature classes or members of a feature dataset.

Soil Conservation Projects

Shoal Lake Study Area

RM Boundaries

Streams

Soil Type

41

94Table 4.1 Topology rules in the geodatabase data model

must be covered by boundary of, must be properly inside

polygons, must be covered by endpoint of, and must be

covered by line

Point

must not overlap, must not intersect, must not have dangles,

must not have pseudo-nodes, must not intersect or touch

interior, must not overlap with, must be covered by feature

class of, must be covered by boundary of, endpoint must be

covered by, must not self overlap, must not self intersect, and

must be single part

Line

must not overlap, must not have gaps, must not overlap with,

must be covered by feature class of, must cover each other,

must be covered by, boundary must be covered by, area

boundary must be covered by boundary of, and contains point

Polygon

RuleFeature

Type

42

10/2/2017

15

ArcGIS Geodatabase Structure

Advantages

Validation rules

– Topology Rules

– Attribute Domains

– Relationship Rules

– Connectivity Rules

– Custom Rules

43

Midterm Exam to Here

Thursday, October 20TH

44

Raster – Vector Conversion

Rasterization

Vectorization

45

10/2/2017

16

115

Figure 5.12

On the left is an example of conversion from vector to raster data, or

rasterization. On the right is an example of conversion from raster to

vector data, or vectorization.

46

Why Convert?

Rasterize

statistical or other mathematical operations

Vectorise

extract thematic information from classified RS

imagery ( roads, rivers, land cover classes)

scanned maps (ArcScan)

47

Why Integrate?

As an image background

Image hotlinks

To subset and classify imagery

Still seldom used in analysis together

48

10/2/2017

17

Rasterization

specify output grid cell size

attribute to be used as cell value

grid cell size v. important

½ min. dimension of smallest feature

inevitable distortion

49

Examples

Straight line rasterization

Bresenham’s Line Algorithm

50

Polygon Rasterisation

How are boundary pixels handled?

Central Point

Dominant Unit

Ranked List

51

10/2/2017

18

Central Point

Dominant Unit

52

Scan Line Coherence

Common polygon rasterisation algorithm

Fill grid cells b/w boundary cells

53

Vectorisation

V. imp for feature extraction

Scanned map products

RS imagery second largest source of geospatial data

54

10/2/2017

19

136

Figure 6.11

Large format drum scanners. (Courtesy of GTCO Calcomp, Inc.)

55

Scanning Maps

Considerations:

Colour depth

Resolution

56

Vectorization

For classified RS images

Reclassification

For scanned maps

Image preprocessing

Thresholding

Bilevel images

57

10/2/2017

20

138

Figure 6.12

A binary scanned

file: the lines are

soil lines, and the

black areas are

the background.

58

Vectorization

Editing/pre-processing boolean image

Remove gaps

Noise

Line thinning

59

139

Figure 6.13

A raster line in a

scanned file has a width

of several pixels.

60

10/2/2017

21

Vectorisation

Vectorization of boolean image

Node detection

Cleaning

Build topology

61